Buccafusco and Heald summarize three arguments put forth by proponents of copyright term extensions. The underuse hypothesis posits that public domain works will be under-exploited; e.g., there is a reduced incentive to make an audiobook of a public domain novel because if it is successful, you cannot prevent copiers from making their own audiobooks and driving down the price. In contrast, the overuse hypothesis posits that public domain works will be over-exploited (like an overplayed song) without owners to maintain their value, so that the audiobook of a public domain novel might be produced so often that it has little value. Finally, according to the misuse hypothesis, the lack of an owner will lead to poor quality or inappropriate uses of public domain works that diminish the value of the underlying work, so that poor-quality audiobooks of public domain novels might cause listeners to think less of the novel itself. To test these hypotheses, Buccafusco and Heald made two lists of bestselling novels: 171 public domain novels published from 1913-22, and 174 copyrighted novels published from 1923-32. From each of these lists, they also made a lists of 20 "durable" novels that are still widely read in addition to being bestsellers when they were published.
Are public domain novels under-exploited or over-exploited in audiobook recordings?
To test the underuse and overuse hypotheses, Buccafusco and Heald searched for audiobook recordings of these novels and found these results:
- 33% of public domain novels are available as audiobooks (3.3 recordings/recorded title; 10% are only available as free LibriVox recordings), compared with 16% of copyrighted novels (3.0 recordings/recorded title). Of the "durable" novels, 100% of public domain novels were available (6.25 recordings/recorded title), compared with 80% of copyrighted novels were available (3.25 recordings/recorded title).
- The average price is similar for professional recordings of public domain and copyrighted novels from the larger lists, but the 20 copyrighted "durable" novels are significantly more expensive than the 20 public domain "durable" novels (e.g., $0.050/min for CDs of copyrighted novel vs. $0.038/min for CDs of public domain novels, meaning the public domain novels are 24% cheaper).
These results illustrate that copyright in the original novel is not necessary to create incentives to make audiobook derivative works; the public domain novels certainly are not being under-exploited. (I was initially concerned that the study might have included audiobooks made before the public domain novels entered the public domain, but the authors have informed me that the only recording in their dataset made before a novel entered the public domain was this 1994 recording of Ulysses, which entered the public domain in 1997.)
But are the public domain novels being over-exploited in audiobooks? Buccafusco and Heald argue no: the public domain novels that were recorded had barely more recordings per title (3.3 vs. 3.0), and even for the durable novels, they argue that 6.25 recordings/title "is not evidence of harmful overuse." I don't know what number of recordings is overuse, but more convincing to me is their pricing data: even though the audiobooks of public domain novels face competition from free versions on LibriVox, the professional versions still command essentially equivalent prices for all novels, and prices that are only 24% cheaper for public domain novels. And the authors argue that even this price difference is more likely caused by not having to pay a copyright royalty and by competition between different editions than by "wearing out" of the novel.
Do poor-quality free recordings of public domain novels tarnish the value of the underlying work?
To test the misuse hypothesis, the authors created a dataset of recordings of their "durable" novels, with 16 recordings in each category: (1) professional copyrighted, (2) professional public domain, (3) amateur copyrighted (recorded by the authors' RA), (4) amateur public domain from LibriVox, and (5) amateur public domain by the authors' RA. Subjects listened to and rated 5-minute recordings on a 1-6 scale (6 = best), and they also indicated how much they thought leftover paperback copies of the work should be sold for (after being told that they typically sell for $8 to $12).
The results are in Table 3, on pages 30-31 of the current draft. Professional recordings of both copyrighted and public domain novels were rated higher than all amateur recordings: professional public domain recordings had the highest average rating of 4.30, followed by professional copyrighted at 4.17; amateur recordings ranged from 3.54 to 3.56. This result refutes the notion that companies might be selecting inferior professional readers for public domain novels. However, because amateur recordings are only publicly available for public domain novels, there will be more poor-quality recordings for these works.
Determining whether these lower-quality recordings might tarnish the underlying work is more difficult. The authors found "a positive and statistically significant relationship between the perceived quality of a recording and the amount that subjects thought copies should be sold for," but "no statistically significant differences in book price between any of the paired conditions" and "no such [statistically significant] correlation within any of the subsamples." (Interestingly, the highest average price of $8.40 was for the RA's recordings of copyrighted works, and the lowest average price of $7.78 was for the RA's recordings of public domain works, but this difference is apparently not statistically significant.)
Buccafusco and Heald conclude that "although there may be a modest feedback effect associated with poor quality versions of creative works, that effect is not related to whether a work is protected by copyright or not." Another possible conclusion is that poor-quality public domain recordings do tarnish the value of the underlying work (as suggested by the statistically significant correlation between quality and price for the whole sample), but that the experiment was unable to resolve this effect. The authors acknowledge this concern: "The failure to find an effect may be the result of a poor experimental design that is not sensitive to differences that actually exist or of insufficient statistical power." But they argue that their "study included hundreds of subjects sourced via multiple methods, and it should have provided the statistical power necessary to detect a difference."
Can we generalize from audiobook recordings of novels to other derivative uses of different copyrighted works?
Buccafusco and Heald convincingly demonstrate that there is still a robust market for high-quality audiobook recordings of public domain novels. But as the authors themselves note, audiobooks are different: "No one is forced to consume an audio book, so multiple copies are not flung in the face of the consuming public who then become tired of hearing the story." They argue that even for music, which appears in commercials and stores, "[m]arket discipline should make over-exploitation highly unlikely" because "businesses try not to alienate their customers by overusing the same music." This conclusion may well be true, but it does not follow from the empirical results of the paper.
The Wikipedia page for the 1998 Copyright Term Extension Act states that an example given by proponents of the act was the film It's a Wonderful Life, which was being "endlessly" broadcast (often as versions that were in "horrible condition") until the owners asserted their rights and gave the film "a high quality restoration that was hailed by critics." One could debate the accuracy of this narrative, but I'm not sure that the audiobook market contributes much to that debate. I am also not convinced that this audiobook study tells us much about amateur fan fiction or trademark law (areas the authors discuss on p. 41).
Ultimately, however, I appreciate that the authors are honest about the limitations of their results, and I agree with their conclusion that the burden of demonstrating harm should lie with proponents of copyright term extensions, especially in light of this study: "[e]ven if works are theoretically harmed by falling into the public domain, proponents of term extension should be expected to establish such losses empirically, because term extension comes with considerable costs [to the public and other creators] that must be justified."
But are the public domain novels being over-exploited in audiobooks? Buccafusco and Heald argue no: the public domain novels that were recorded had barely more recordings per title (3.3 vs. 3.0), and even for the durable novels, they argue that 6.25 recordings/title "is not evidence of harmful overuse." I don't know what number of recordings is overuse, but more convincing to me is their pricing data: even though the audiobooks of public domain novels face competition from free versions on LibriVox, the professional versions still command essentially equivalent prices for all novels, and prices that are only 24% cheaper for public domain novels. And the authors argue that even this price difference is more likely caused by not having to pay a copyright royalty and by competition between different editions than by "wearing out" of the novel.
Do poor-quality free recordings of public domain novels tarnish the value of the underlying work?
To test the misuse hypothesis, the authors created a dataset of recordings of their "durable" novels, with 16 recordings in each category: (1) professional copyrighted, (2) professional public domain, (3) amateur copyrighted (recorded by the authors' RA), (4) amateur public domain from LibriVox, and (5) amateur public domain by the authors' RA. Subjects listened to and rated 5-minute recordings on a 1-6 scale (6 = best), and they also indicated how much they thought leftover paperback copies of the work should be sold for (after being told that they typically sell for $8 to $12).
The results are in Table 3, on pages 30-31 of the current draft. Professional recordings of both copyrighted and public domain novels were rated higher than all amateur recordings: professional public domain recordings had the highest average rating of 4.30, followed by professional copyrighted at 4.17; amateur recordings ranged from 3.54 to 3.56. This result refutes the notion that companies might be selecting inferior professional readers for public domain novels. However, because amateur recordings are only publicly available for public domain novels, there will be more poor-quality recordings for these works.
Determining whether these lower-quality recordings might tarnish the underlying work is more difficult. The authors found "a positive and statistically significant relationship between the perceived quality of a recording and the amount that subjects thought copies should be sold for," but "no statistically significant differences in book price between any of the paired conditions" and "no such [statistically significant] correlation within any of the subsamples." (Interestingly, the highest average price of $8.40 was for the RA's recordings of copyrighted works, and the lowest average price of $7.78 was for the RA's recordings of public domain works, but this difference is apparently not statistically significant.)
Buccafusco and Heald conclude that "although there may be a modest feedback effect associated with poor quality versions of creative works, that effect is not related to whether a work is protected by copyright or not." Another possible conclusion is that poor-quality public domain recordings do tarnish the value of the underlying work (as suggested by the statistically significant correlation between quality and price for the whole sample), but that the experiment was unable to resolve this effect. The authors acknowledge this concern: "The failure to find an effect may be the result of a poor experimental design that is not sensitive to differences that actually exist or of insufficient statistical power." But they argue that their "study included hundreds of subjects sourced via multiple methods, and it should have provided the statistical power necessary to detect a difference."
Can we generalize from audiobook recordings of novels to other derivative uses of different copyrighted works?
Buccafusco and Heald convincingly demonstrate that there is still a robust market for high-quality audiobook recordings of public domain novels. But as the authors themselves note, audiobooks are different: "No one is forced to consume an audio book, so multiple copies are not flung in the face of the consuming public who then become tired of hearing the story." They argue that even for music, which appears in commercials and stores, "[m]arket discipline should make over-exploitation highly unlikely" because "businesses try not to alienate their customers by overusing the same music." This conclusion may well be true, but it does not follow from the empirical results of the paper.
The Wikipedia page for the 1998 Copyright Term Extension Act states that an example given by proponents of the act was the film It's a Wonderful Life, which was being "endlessly" broadcast (often as versions that were in "horrible condition") until the owners asserted their rights and gave the film "a high quality restoration that was hailed by critics." One could debate the accuracy of this narrative, but I'm not sure that the audiobook market contributes much to that debate. I am also not convinced that this audiobook study tells us much about amateur fan fiction or trademark law (areas the authors discuss on p. 41).
Ultimately, however, I appreciate that the authors are honest about the limitations of their results, and I agree with their conclusion that the burden of demonstrating harm should lie with proponents of copyright term extensions, especially in light of this study: "[e]ven if works are theoretically harmed by falling into the public domain, proponents of term extension should be expected to establish such losses empirically, because term extension comes with considerable costs [to the public and other creators] that must be justified."
